Multi-factor pairing identification system for multiple vechicle characteristics, method and computer readable medium thereof

Abstract

This invention relates to a multi-factor pairing and identification system and method for vehicles with multiple features. It utilizes various identification devices to collect multiple vehicle feature information, and through a machine learning model, pairs and integrates the vehicle identification feature information from multiple devices into a single vehicle passage information to achieve accurate vehicle identification. Based on this, the invention can improve the problem of image information identification errors caused by environmental factors, as well as vehicle pairing information errors caused by multi-lane or high-traffic environments. Furthermore, it incorporates a vehicle identification learning database to correct vehicle identification results, improving the overall vehicle identification accuracy. This invention also provides a computer-readable medium for performing the method of this invention.

Description

Technical Field

[0001] This invention relates to a multi-feature identification and pairing technology, and in particular to a multi-factor pairing identification system, method and computer-readable medium for vehicle multi-features. Prior Technology

[0002] Currently, vehicle recognition systems are used for toll collection and access control in many venues. A common and conventional technology is image recognition to capture vehicle license plate information. However, images are easily affected by environmental factors such as weather and damaged license plates, leading to recognition rate issues. When a vehicle recognition system used for access control makes a judgment error or a mismatch in information, vehicles may be unable to pass, requiring manual intervention. In situations with high traffic volume, this can cause lane congestion, poor traffic efficiency, and prolonged idling for vehicles behind, resulting in air pollution and energy waste.

[0003] Another known technology involves pairing a license plate with a third-party vehicle identification tag and using the cumulative number of passes to identify the vehicle. This technology uses the time the vehicle passes as the pairing condition and integrates the two types of information into a single vehicle information. However, when this technology is used in multi-lane or high-traffic environments, there is still a chance of pairing errors due to radio frequency signal problems, such as mispairing to other lanes or preceding or following vehicles.

[0004] Another known technique is to use a single coil trigger signal to perform asynchronous pairing of third-party vehicle tags and license plate numbers. Although this technique can perform vehicle feature pairing with higher accuracy, it still has the limitation of requiring a single coil trigger signal source, as well as the problem of misreading the radio frequency signal of the third-party vehicle tag and identifying the radio frequency tags of adjacent lanes or vehicles in front and behind, which leads to the probability of pairing errors.

[0005] It is evident that the aforementioned conventional methods still have many shortcomings and are not a good design, and urgently need to be improved. Therefore, how to find a matching technology for multiple features, especially the integration of multiple parameter factors to match vehicle feature information, and thus achieve the purpose of vehicle identification, has become the goal that people in this technical field are eager to pursue. Summary of the Invention

[0006] To achieve the aforementioned objectives, this invention proposes a multi-factor pairing and identification system for vehicles with multiple features, comprising: an image recognition module for acquiring vehicle identification information; a third-party tag reading module for acquiring third-party tag information of the vehicle; a multi-factor pairing module connected to the image recognition module and the third-party tag reading module for receiving the identification information and the third-party tag information, generating a pairing score through a machine learning model, and determining the pairing result based on the pairing score; a vehicle identification learning database for storing previous pairing data of vehicle license plates and third-party tag information, as well as the cumulative pairing learning score; and a vehicle identification learning module connected to the multi-factor pairing module for comparing the vehicle identification learning database with the pairing data and the cumulative pairing learning score based on the pairing result provided by the multi-factor pairing module, thereby searching for matching data and correcting the license plate.

[0007] In one embodiment, the image recognition module is a plurality of image recognition cameras used to capture the license plate image of the vehicle as the recognition information.

[0008] In one embodiment, the third-party tag reading module is used to read third-party tags installed on the vehicle to obtain third-party tag information, wherein the third-party tag is a radio frequency tag (RFID), an on-board unit (OBU), or a ZigBee tag.

[0009] This invention proposes a multi-factor pairing method for vehicle identification, comprising the following steps: having an image recognition module and a third-party tag reading module obtain vehicle identification information and third-party tag information; and having the multi-factor pairing module generate a pairing score using the identification information and the third-party tag information through a machine learning model, wherein when the pairing score is greater than a pairing threshold and is the maximum value, the pairing data corresponding to the pairing score is taken as the pairing result for this time.

[0010] In the above method, the recognition information obtained by the image recognition module is the license plate number, vehicle passage time, or license plate recognition quality index.

[0011] In the above method, the third-party tag information obtained by the third-party tag reading module is the time, signal strength, or antenna number of the third-party tag of the vehicle.

[0012] In the above method, the pairing threshold value for the pairing score is a parameter preset by the system.

[0013] The present invention further discloses a computer-readable medium, which is used in a computing device or computer, and stores instructions to execute the aforementioned multi-factor matching method for vehicle identification.

[0014] This invention proposes a multi-factor pairing learning method for vehicle recognition, comprising the following steps: a multi-factor pairing module obtains vehicle pairing results through a machine learning model; after obtaining the pairing results from the multi-factor pairing module, the vehicle recognition learning module writes the pairing data of the pairing results into a vehicle recognition learning database, searches historical pairing results in the vehicle recognition learning database, and finds data that matches and whose cumulative learning score is greater than a cumulative threshold value, as a reference for license plate correction; and the vehicle recognition learning module determines the number of codes for license plate correction based on the confidence level of the image recognition module in recognizing the license plate and the pairing score generated by the multi-factor pairing module.

[0015] In the above method, the determination of the license plate correction code includes the license plate recognition quality index.

[0016] In the above method, the cumulative threshold value for the cumulative learning score is a parameter preset by the system.

[0017] The present invention further discloses a computer-readable medium, which is used in a computing device or computer, and stores instructions to execute the aforementioned multi-factor pairing learning method for vehicle recognition.

[0018] In summary, the multi-factor pairing identification system, multi-factor pairing method, and multi-factor pairing learning method of the present invention aim to provide a method for multi-factor pairing and identification using multiple vehicle features. Specifically, in the vehicle identification lane, an image identification module and a third-party tag reading module are combined to identify vehicle information. The different types of information features provided by each module are paired and integrated into a single vehicle number through a machine learning model, such as a multi-factor pairing model. This improves the problem of incorrect pairing of multiple feature information in the past. In addition, the present invention incorporates a vehicle identification learning database to correct vehicle identification data, which will help improve the identification accuracy of the vehicle identification system. Simple Explanation of the Diagram

[0019] Please refer to the detailed description and accompanying drawings of this invention for a further understanding of its technical content and objectives. The accompanying drawings are as follows:

[0020] Figure 1 is a system architecture diagram of the multi-factor pairing identification system for multiple vehicle features of the present invention.

[0021] Figure 2 is a flowchart illustrating the steps of the multi-factor pairing method for vehicle identification according to the present invention.

[0022] Figure 3 is a flowchart illustrating the steps of the multi-factor pairing learning method for vehicle identification according to the present invention.

[0023] Figure 4 is a flowchart of the multi-factor pairing identification method for multiple features of vehicles according to the present invention. Implementation

[0024] The technical content of this invention is described below through specific embodiments. Those skilled in the art can easily understand the advantages and effects of this invention from the content disclosed in this specification. However, this invention can also be implemented or applied through other different embodiments.

[0025] Figure 1 is a system architecture diagram of the multi-factor pairing and identification system for vehicle multiple features of the present invention. The purpose of the present invention is to use vehicle multiple features for multi-factor pairing and identification. As shown in the figure, the multi-factor pairing and identification system 1 for vehicle multiple features of the present invention includes an image recognition module 11, a third-party tag reading module 12, a multi-factor pairing module 13, a vehicle identification learning database 14, and a vehicle identification learning module 15.

[0026] The image recognition module 11 is used to acquire vehicle identification information, and the third-party tag reading module 12 is used to acquire third-party tag information of the vehicle. The multi-factor pairing module 13 connects the image recognition module 11 and the third-party tag reading module 12. Specifically, the image recognition module 11 is a plurality of image recognition cameras, such as an optical character recognition (OCR) image recognition camera, used to capture the license plate image of the vehicle as the identification information. That is, the purpose of the image recognition module 11 is to acquire image recognition information. In addition, the third-party tag reading module 12 is used to continuously read third-party tag information, that is, to read the third-party tags installed on the vehicle to become the third-party tag information. The third-party tags can be radio frequency tags (RFID), on-board units (OBU), and / or ZigBee tags.

[0027] The multi-factor matching module 13 receives recognition information from the image recognition module 11 and third-party tag information from the third-party tag reading module 12. It generates a matching score by using a machine learning model to process the recognition information and third-party tag information, and then determines the matching result of this vehicle based on the matching score. The vehicle recognition learning database 14 stores the matching data and cumulative matching learning score of the license plate and third-party tag information of previous vehicles. The vehicle recognition learning module 15 is connected to the multi-factor matching module 13 and uses the matching result provided by the multi-factor matching module 13 to compare the vehicle recognition learning database 14 with the matching data and the cumulative matching learning score, thereby searching for matching data to correct the license plate.

[0028] In one embodiment, the present invention utilizes multiple identification devices to collect multiple vehicle feature information. Through a machine learning model, the vehicle identification feature information from multiple devices is paired and integrated into a single vehicle passage information, thus achieving the purpose of accurate vehicle identification. Specifically, the image recognition module 11 uses image recognition to identify the identification information of passing vehicles, and the third-party tag reading module 12 can read the third-party vehicle tags of the vehicles to obtain third-party tag information. The above information is transmitted to the multi-factor pairing module 13. The multi-factor pairing module 13 uses a machine learning model to pair the identification information with the third-party tag information and other feature factor information to integrate the vehicle information into a single vehicle passage information. Afterwards, the vehicle identification learning module 15 can read and write the vehicle identification learning database 14 and compare and correct the license plate, thus providing the corrected license plate identification result. The vehicle learning database 14 is then used to accumulate and learn vehicle passage pairing information. Based on the above, the present invention utilizes the image recognition module 11 to obtain image recognition information, and combines it with the third-party tag reading module 12 to obtain third-party tag information. Through the multi-factor matching module 13, the image recognition information and the third-party tag information are matched by a machine learning model to integrate them into a single vehicle information, which is then written into the vehicle recognition learning database 14 to correct the vehicle recognition result. Therefore, the present invention can optimize the vehicle recognition result.

[0029] In summary, the multi-factor matching and identification system 1 for vehicles with multiple features of the present invention can improve the problem of image information identification errors caused by environmental factors. By adding the reading results of third-party vehicle tags and then matching and integrating vehicle information through the multi-factor matching module, the problem of vehicle matching information errors caused by multi-lane or high traffic flow environments can be improved. Therefore, the accuracy of vehicle information matching obtained by each module can be improved. In addition, the present invention also combines a vehicle identification learning database to correct vehicle identification results, which also helps to improve the overall vehicle identification accuracy.

[0030] In one embodiment, each module of the present invention can be software, hardware, or firmware; if it is hardware, it can be a processing unit, processor, computer, or server with data processing and computing capabilities; if it is software or firmware, it can include instructions executable by the processing unit, processor, computer, or server, and can be installed on the same hardware device or distributed across different multiple hardware devices.

[0031] Figure 2 is a flowchart illustrating the steps of the multi-factor pairing method for vehicle identification according to the present invention. As shown in the figure, the present invention obtains multiple features of a vehicle to perform multi-factor pairing.

[0032] In step S201, the image recognition module and the third-party tag reading module obtain vehicle identification information and third-party tag information. This step explains that the image recognition module obtains vehicle identification information, and the third-party tag reading module obtains third-party tag information for the vehicle. This information will be integrated into the information for that vehicle.

[0033] In one embodiment, the recognition information obtained by the image recognition module may be the license plate number, vehicle passage time, and / or license plate recognition quality index, but is not limited thereto.

[0034] In one embodiment, the third-party tag information obtained by the third-party tag reading module may be the time of reading the third-party tag of the vehicle, the signal strength, and / or the reading antenna number, but is not limited thereto.

[0035] In step S202, the multi-factor matching module uses a machine learning model to generate a matching score based on the identification information and the third-party label information. When the matching score is greater than a matching threshold and is the maximum value, the matching data corresponding to that matching score is taken as the matching result for this time. This step explains that the multi-factor matching module uses the obtained identification information and third-party label information to generate a matching score through a machine learning model. Then, it compares the matching score with a matching threshold. When the matching score is greater than the matching threshold and is the maximum value (because there may be multiple matching scores), the matching data corresponding to that matching score becomes the matching result for this time.

[0036] In one embodiment, the pairing threshold value for the pairing score is a preset parameter of the system, i.e., a preset value within the system.

[0037] Figure 3 is a flowchart illustrating the steps of the multi-factor pairing learning method applied to vehicle identification according to the present invention. As shown in the figure, the present invention performs multi-factor pairing learning based on multiple features of the vehicle.

[0038] In step S301, the multi-factor matching module obtains the vehicle matching result through a machine learning model. This step explains that the multi-factor matching module can obtain the vehicle matching result by using the vehicle identification information and third-party tag information through a machine learning model.

[0039] In step S302, after obtaining the pairing result from the multi-factor pairing module, the vehicle recognition learning module writes the pairing data of the pairing result into the vehicle recognition learning database. It then searches the database for historical pairing results to find data that matches the pairing and whose cumulative learning score is greater than a cumulative threshold. This data serves as a reference for correcting the vehicle's license plate. This step explains that after obtaining the pairing result from the aforementioned step, the vehicle recognition learning module compares the pairing data with the vehicle recognition learning database to search for historical pairing results. It then finds data that matches the pairing and whose cumulative learning score is greater than a cumulative threshold. This data serves as a reference for correcting the vehicle's license plate. In other words, by comparing the pairing result with historical data, it finds information that matches the current pairing result. When this information meets the cumulative threshold, it indicates that the information is more suitable for the vehicle's recognition, and therefore, the vehicle's license plate is corrected accordingly.

[0040] In step S303, the vehicle recognition learning module determines the license plate correction code based on the confidence level of the image recognition module in recognizing the license plate and the matching score generated by the multi-factor matching module. This step explains the method of license plate correction, that is, the license plate correction can be based on the vehicle identification-related information, which may include the confidence level of the image recognition module in recognizing the license plate and the matching score generated by the multi-factor matching module. The above information can be used as a reference for the license plate correction code.

[0041] In one embodiment, the basis for determining the number of codes for license plate correction may also include a license plate recognition quality index.

[0042] In one embodiment, the cumulative threshold value for the cumulative learning score is a parameter preset by the system, that is, a parameter value pre-set within the system.

[0043] Figure 4 is a flowchart of the multi-factor pairing identification method for multiple features of vehicles according to the present invention.

[0044] In process 401, the image recognition module obtains the license plate information of the passing vehicles. This process involves the image recognition module obtaining the license plate information of the passing vehicles from the captured images, which may include the vehicle passage time PT and the license plate recognition confidence level PV.

[0045] In process 402, the third-party tag reading module obtains third-party tag information. This process describes how the third-party tag reading module continuously reads and obtains one or more third-party vehicle tag Ii, reading time RTi, reading antenna number Ai, and reading signal strength Ri, where i is the read third-party vehicle tag number, and 0 is the third-party vehicle tag number. i Nread, where Nread represents the number of third-party vehicle tags read.

[0046] In process 403, the multi-factor pairing module obtains the information provided by the aforementioned module and converts it into feature data such as reading time difference and signal strength required by the machine learning model. This process explains that the multi-factor pairing module integrates the time information of the image recognition module and the time information RTi and reading signal strength Ri of the third-party tag reading module, and converts them into input features of the machine learning model. The features include, but are not limited to, (1) the antenna number Ai of the i-th third-party tag received, (2) the signal strength Ri of the i-th third-party tag received, (3) the time difference PT-RTi between the vehicle's head passing and the earliest reading of the i-th third-party tag, (4) the time difference PT-RTN between the vehicle's head passing and the latest reading of the third-party tag, and (5) the number of times the i-th third-party tag is read Nread,I within the time from the 30 seconds before the vehicle's head passes to the vehicle's departure.

[0047] In process 404, the multi-factor matching module uses a machine learning model to predict the matching score between the third-party label and the license plate based on the transformed features. This process explains that the multi-factor matching module uses a machine learning model to predict the matching score (MV) between the license plate and the third-party label information based on the transformed various feature information. The machine learning model for multi-factor matching uses previously confirmed vehicle passage records, the identification results of various vehicle identification devices, and their factors to build the model, and can continuously collect and optimize the model according to the environment of different project sites.

[0048] Specifically, depending on the different case environment, the multi-factor pairing machine learning model architecture can be implemented as XGBoost (eXtreme Gradient Boosting), Random Forest, LightGBM (Light Gradient Boosting Machine), Multilayer Perceptron (MLP), etc., but the present invention is not limited to these.

[0049] In process 405, determine whether the pairing score is greater than a set threshold value. This process sets a pairing score threshold value TV, determines whether the pairing score TW is greater than the score threshold value TV, if not, returns to process 404, and predicts the pairing score again, if yes, proceeds to process 406.

[0050] In process 406, the multi-factor matching module obtains the matching information with the highest matching score as the matching result for that vehicle trip. This process explains that a single vehicle passage record may have multiple sets of third-party vehicle tag matching scores MVj that meet a set threshold, where j is the third-party vehicle tag number whose matching score is higher than the set threshold, where 0... j Nmapping, where Nmapping is the number of third-party vehicle tag matching scores that are higher than a set threshold. The matching data with the highest matching score MVmax is the matching result for that vehicle.

[0051] In process 407, the vehicle recognition learning module searches the vehicle recognition learning database for information that matches the matching criteria. This process explains that the vehicle recognition learning module will search the license plate recognition learning database for data that matches the matching results from process 406.

[0052] In process 408, it is determined whether the accumulated learning score is greater than the set threshold value. This process determines whether the accumulated learning score is greater than the set cumulative threshold value. If not, it returns to process 407 to search for suitable matching data again. If so, it means that the data can be used as a correction reference for this train, and it will proceed to process 409.

[0053] In process 409, the vehicle recognition learning module corrects the license plate. If no correction is made, the matching result is written to the vehicle recognition learning database to continuously accumulate learning scores. This process corrects the license plate, and the number of correction codes can be determined based on the confidence level of the image recognition module in recognizing the license plate and the matching score of the multi-factor matching module. If the difference between the license plate code and the number of correction codes obtained from the matching database is too large, no correction is made.

[0054] In process 410, the vehicle recognition learning module obtains the vehicle recognition result. This process describes how the vehicle recognition learning module obtains the vehicle recognition result for this particular train.

[0055] Furthermore, the present invention discloses a computer-readable medium applied in a computing device or computer having a processor (e.g., CPU, GPU, etc.) and / or memory, storing instructions, and capable of being executed by the computing device or computer through the processor and / or memory to perform the aforementioned methods and steps or processes when executing the computer-readable medium. In one embodiment, the computer-readable medium is a non-transitory computer-readable storage medium.

[0056] In summary, the multi-factor pairing identification system, multi-factor pairing method, and multi-factor pairing learning method of this invention utilize image recognition to capture vehicle information, combine it with reading third-party vehicle tags, and integrate the results transmitted by each module into single vehicle information through multi-factor pairing. By combining this with a vehicle identification learning database to correct the vehicle identification results, the invention can improve the problem of image information identification errors caused by environmental factors. By incorporating third-party vehicle tag reading results and then using a multi-factor pairing module to pair and integrate vehicle information, the invention improves the vehicle pairing information error problem caused by multi-lane or high-traffic environments, increases the accuracy of vehicle information pairing obtained by each module, and improves the overall vehicle identification accuracy by combining the vehicle identification learning database to correct the vehicle identification results.

[0057] Compared with other conventional technologies, the present invention has the following advantages:

[0058] First, this invention utilizes a machine learning model to establish a multi-factor pairing module. By referencing the feature information of different types of identification modules, it selects combinations with higher predicted pairing probabilities, effectively reducing the probability of pairing errors.

[0059] Secondly, this invention improves the overall vehicle recognition system's accuracy by combining image recognition information and third-party tag information with a multi-factor pairing module and a learning database of accumulated pairing information to correct the recognition results.

[0060] The above detailed description is a specific description of one feasible embodiment of the present invention. However, this embodiment is not intended to limit the patent scope of the present invention. All equivalent implementations or modifications that do not depart from the spirit of the present invention should be included in the patent scope of the present invention.

[0061]

[0062] 1: Multi-factor pairing identification system for multiple vehicle features

[0063] 11: Image Recognition Module

[0064] 12: Third-party tag reading module

[0065] 13: Multi-factor pairing module

[0066] 14: Vehicle Recognition Learning Resource Database

[0067] 15: Vehicle Recognition Learning Module

[0068] 401-410: Process

[0069] S201-S202: Steps

[0070] S301-S303: Steps

Claims

1. A multi-factor pairing identification system for multiple vehicle features, comprising: Image recognition module, used to obtain vehicle identification information; A third-party tag reading module is used to obtain the third-party tag information of the vehicle; The multi-factor matching module connects the image recognition module and the third-party tag reading module to receive the recognition information and the third-party tag information. It generates a matching score through a machine learning model and determines the matching result based on the matching score. The vehicle recognition learning database stores the matching data and cumulative matching learning scores of the license plates of previous vehicles with third-party tag information. The vehicle recognition learning module connects to the multi-factor matching module and compares the matching data and the cumulative matching learning scores with the matching results provided by the multi-factor matching module. It searches for matching data whose cumulative matching learning scores are greater than a cumulative threshold value, which serves as a reference for correcting the vehicle's license plate. Based on the confidence level of the image recognition module in recognizing the license plate and the matching score generated by the multi-factor matching module, the number of codes for license plate correction is determined.

2. A multi-factor pairing identification system for multiple vehicle features as described in claim 1, wherein, The image recognition module consists of multiple image recognition cameras used to capture the license plate image of the vehicle as the recognition information.

3. A multi-factor pairing identification system for multiple vehicle features as described in claim 1, wherein, The third-party tag reading module is used to read third-party tags installed on the vehicle to obtain third-party tag information. The third-party tag is a radio frequency tag (RFID), an on-board unit (OBU), or a ZigBee tag.

4. A multi-factor pairing identification system for multiple vehicle features as described in claim 1, wherein, When the pairing score is greater than a pairing threshold and is the maximum value, the multi-factor pairing module uses the pairing data corresponding to the pairing score as the pairing result for this time.

5. A multi-factor pairing identification system for multiple vehicle features as described in claim 4, wherein, The image recognition module obtains the license plate number, vehicle passage time, or license plate recognition quality index.

6. A multi-factor pairing identification system for multiple vehicle features as described in claim 4, wherein, The third-party tag reading module obtains information about the third-party tag, including the time, signal strength, or antenna number of the third-party tag on the vehicle.

7. A multi-factor pairing identification system for multiple vehicle features as described in claim 4, wherein, The pairing threshold value for this pairing score is a system-preset parameter.

8. A multi-factor pairing learning method for vehicle recognition, comprising the following steps: having a multi-factor pairing module obtain vehicle pairing results through a machine learning model; having a vehicle recognition learning module, after obtaining the pairing results from the multi-factor pairing module, write the pairing data of the pairing results into a vehicle recognition learning database, thereby searching for historical pairing results from the vehicle recognition learning database, and finding data that matches and whose cumulative learning score is greater than a cumulative threshold value, as a reference for license plate correction of the vehicle; and having the vehicle recognition learning module determine the number of codes for license plate correction based on the confidence level of the license plate recognition by the image recognition module and the pairing score generated by the multi-factor pairing module.

9. The multi-factor pairing learning method for vehicle identification as described in claim 8, wherein, The determination of the corrected license plate code includes the license plate recognition quality index.

10. The multi-factor pairing learning method for vehicle identification as described in claim 8, wherein, The cumulative threshold value for the accumulated learning score is a parameter preset by the system.

11. A computer-readable medium, applied in a computing device or computer, storing instructions for performing a multi-factor pairing learning method for vehicle identification as described in any one of claims 8 to 10.